US5724594A - Method and system for automatically identifying morphological information from a machine-readable dictionary - Google Patents
Method and system for automatically identifying morphological information from a machine-readable dictionary Download PDFInfo
- Publication number
- US5724594A US5724594A US08/195,346 US19534694A US5724594A US 5724594 A US5724594 A US 5724594A US 19534694 A US19534694 A US 19534694A US 5724594 A US5724594 A US 5724594A
- Authority
- US
- United States
- Prior art keywords
- word
- derived
- morpheme
- semantic
- semantic information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Lifetime
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/247—Thesauruses; Synonyms
Definitions
- This invention relates generally to a computer system for deriving semantic information from machine-readable dictionaries, and more specifically, to a method and system for identifying morphologically related words in machine-readable dictionaries.
- Machine-readable dictionaries often contain definitional information for each word in such a way that that information is accessible by a computer program.
- MRDs typically contain explicit and complete information making it possible to identify inflected forms of words. That information may appear in the form of orthographic rules or as a list of regular and irregular forms.
- Regular inflectional forms can be derived from rules in the "Guide to the Dictionary,” which is typically included with the dictionary. Irregular inflectional forms are explicitly listed in dictionary entries. For example, the entry for the verb "sing" lists the irregular forms “sang” and “sung” and the regular forms “sings” and “singing.”
- MRDs typically contain only implicit and scant information or word formation (derivation). Thus, it is difficult for even a human user to construct a list of all words derived from a given word. For example, the following words, which are morphologically related to the verb "believe,” are defined separately in typical MRDs: “belief,” “believe,” “believable,” “unbelievable,” “believer,” “disbelief,” “disbelieve,” “unbelief,” “unbelieving,” and “unbelievingly.” These words are not marked explicitly in the dictionary as being morphologically related. The word “believe” is the base word of these derived words.
- the word “believer” is formed from the base word “believe” and the morpheme "er.”
- a morpheme is a meaningful linguistic unit that cannot be further divided into meaningful parts.
- Automated methods have been developed for parsing a derived word into its component parts: putative base word and morpheme. Generally, these methods use finite state morphological analysis as described in Sproat, R., "Morphology and Computation," The MIT Press, 1992. Finite state morphological analysis applies morpheme-based rules to identify the component parts. For example, one rule may be that the morpheme "ize” attached to a noun or an adjective creates a verb.
- the word “hospitalize” can be parsed into “hospital” and “ize.”
- the word “maize” would not parse into “ma” and “ize,” because “maize” is a noun and not a verb.
- successful parsing of a word does not guarantee a morphological relationship between the word and its putative base word.
- the word “flower” may parse into “flow” and “er,” but the word “flower” is not morphologically related to the word “flow.”
- MRDs may include some form of overt markings of morphological relatedness. Specifically, some undefined forms may be listed at the end of the dictionary entry for the word from which they are ultimately derived. The MRDs assume that the meaning of these words should be clear when the meaning of the suffix is added to the meaning of the base word. For example, the dictionary entry for the word “journalism” may indicate that the form “journalistic” is derived from that word.
- the present invention provides a method and system for determining the derivational relatedness of a derived word and a base word.
- the system includes a database with an entry for each head word and morpheme. Each entry contains definitional information from a machine-readable dictionary and semantic relations extracted from the definitions. Each semantic relation specifies a relation between the head word with a word used in its definition. Semantic relations may contain nested semantic relations to specify relations between words in the definition.
- the system identifies a putative base word of a derived word preferably using finite state morphological analysis.
- the system compares the semantic relations of the derived word to the semantic relations of the morpheme, which is augmented with the base word.
- the system generates a derivational score that indicates the confidence that the derived word derive from the base word.
- FIG. 1 is a block diagram of a computer system of a preferred embodiment.
- FIG. 2 is a flow diagram of a function CompareContent.
- the present invention provides a system and method for automatically identifying morphologically related words.
- the system known as Morels for morphological relation system, determines whether a word, known as a derived word, is indeed derived from a putative base word identified preferably using finite state morphological analysis.
- the system assigns a derivational score to the putative derivation indicating the likelihood that the derived word is a derivative of the base word.
- the system preferably uses a database containing semantic relations for each head word and morpheme in a machine-readable dictionary (MRD).
- MRD machine-readable dictionary
- a head word is a word for which the dictionary contains an entry.
- a morpheme is a meaningful linguistic unit that cannot be divided into further meaningful parts (e.g., "er,” “ed,” “ing,” and “ion”).
- a derived word is considered to be a combination of the putative base word and a morpheme.
- the system compares the semantic relations of the morpheme augmented with the base word to the semantic relations for the derived word.
- a semantic relation e.g., genus, subject, object
- the semantic relations are generated automatically by parsing the definition of a head word. Based on the comparison, the system assigns a derivational score to the putative derivation. This derivational score indicates the likelihood that the derived word derives from the putative base word.
- FIG. 1 is a block diagram of a computer system of a preferred embodiment.
- the computer system includes a central processing unit 101, a memory 102, a database 103, and a display 104.
- the memory contains executable computer code 105 implementing a preferred embodiment of the present invention.
- the database contains a dictionary with semantic relations specified.
- the database contains an entry for each head word and morpheme of a dictionary.
- Each head word entry contains a list of semantic relations for each sense (different definition) of the head word.
- Each semantic relation comprises an attribute (type) and has as its value a list of one or more semantic records.
- a semantic record contains a lemma value and optionally one or more nested semantic relations.
- a lemma value is the uninflected form of the word, which is typically the form under which the word is listed in the dictionary as a head word. For example, the lemma of "takes,” “taking,” “took,” and “taken” is the form “take” and the lemma of "children” is
- Table 1 contains the entry for the head word "banker.”
- the entry contains two definitions or senses: (1) "a person who owns or controls or shares in the control of a bank” and (2) "the player who keeps the bank in various games of chance.”
- Each sense contains the definition of the sense and the list of semantic relations for that sense.
- the first sense contains two semantic relations.
- the first semantic relation is
- the semantic relation type is "Hypernym,” and its value is the semantic record “ ⁇ Lemma "person” ⁇ .”
- the lemma value is "person,” and the semantic record contains no nested semantic relations.
- the second semantic relation is
- the semantic relation type is "SubjOf,” and its value is a list of three semantic records. Each semantic record contains a nested semantic relation with the semantic relation type of "HasObj.”
- An unnested semantic relation relates to the head word (the related-to word).
- a nested semantic relation relates to the lemma value (the related-to word) of the nesting semantic record.
- a "Hypernym” type identifies that the lemma value is the genus for which the related to word is a species.
- a "SubjOf” type identifies that the related-to word is the subject of the lemma value.
- a “HasObj” type identifies that the lemma value is the object of the related-to word.
- a "PartOf” type identifies that the related-to word is part of the lemma value.
- the semantic relations for the first sense of "banker” indicate that "banker” is a kind of “person” (Hypernym), that "banker” is the subject of “own” (SubjOf, and that "bank” is the object of "own” (HasObj).
- Table 2 contains examples illustrating the meaning of various types of semantic relations.
- Table 3 contains the morpheme entry for the morpheme
- Morpheme semantic relations contain Morels values.
- a Morels value indicates the amount by which a derivational score is to be increased when structural isomorphism and equivalent lemma values are encountered during the evaluation of morphological relatedness between the derived word and the putative base word.
- Structural isomorphism occurs when words have similar hierarchy of semantic relation types.
- a head word entry with a "Hypernym” and “SubjOf” semantic relation and with a “HasObj” semantic relation nested within the "SubjOf” semantic relation is structurally isomorphic to the morpheme "er.”
- the lemma value of "base word” indicates that the putative base word should be inserted to complete the semantic relations for the putative base word combined with the morpheme.
- the lemma value for a morpheme may be represented by a list of lemma values ("know", "work”) when their semantic records are structurally isomorphic with equivalent lemma values.
- the semantic relations for each head word of dictionary database may be generated manually or automatically.
- the system uses a database generated as described by Dolan, Vander Liste, and Richardson in Proceedings of the First Conference of the Pacific Association for Computational Linguistics, 1993, pp. 5-14; Jensen and Binot, "A Semantic Expert Using On-Line Standard Dictionary,” reprinted in Natural Language Processing: The PLNLP Approach, 1992, chap. 11; and Montemagni and Vander Liste, Proceedings of Coling, 1992, pp. 546-52, all of which are hereby incorporated by reference. Since the number of morphemes in a dictionary is relatively small, the Morels values and the morpheme semantic relations are preferably manually entered into the database.
- the preferred embodiment is described using a MRD with semantic relations already specified, one skilled in the art would appreciate that the semantic relation information can be dynamically generated during derivational analysis from definitional information in the MRD.
- the system generates a derivational score for a sense of a derived word that indicates the likelihood that the derived word is a derivation of a putative base word and a morpheme.
- the system generates an instance of the morpheme semantic relations inserting the putative base word where indicated.
- the system selects each semantic relation of the instantiated morpheme and determines whether the derived word has a semantic relation of the same type. If the derived word has such a semantic relation and the lemma value of the instantiated morpheme and the derived word are equal, then the system increases the derivational score.
- the system determines whether the lemma value of the derived word is related to the lemma value of the instantiated morpheme. If the semantic relation type is not "Hypernym,” then the system determines if the base word is related to the lemma value of the derived word. The system then recursively determines whether nested semantic relations of the instantiated morpheme are related to the nested semantic relations of the derived word.
- FIG. 2 is a flow diagram illustrating a CompareContent function of the system.
- the function is passed a list of semantic relations for a morpheme (LSR m ) and a list of semantic relations for a derived word (LSR d ).
- the function determines derivational relatedness of the derived word and the putative base word.
- the function recursively invokes itself to process nested semantic relations.
- the function is initially invoked with a global derivational score set to zero.
- the function increases the derivational score to indicate derivational relatedness.
- the function selects the next semantic relation (SR m ) of the morpheme starting with the first semantic relation in LSR m .
- the function loops processing each semantic relation in the morpheme.
- step 202 if all the semantic relations for the morpheme have already been selected, then the function returns, else the function continues at step 203.
- step 203 if the list of semantic relations (LSR d ) for the derived word contains a semantic relation (SR d ) with a type equal to the selected semantic relation (SR m ) of the morpheme, then the function continues at step 204, else the function loops to step 201 to select the next semantic relation (SR m ) of the morpheme.
- step 204 the function selects the next semantic record (REC d ) of the selected semantic relation (SR d ) of the derived word starting with the first semantic record.
- step 205 if all semantic records (REC d ) have already been selected, then the function loops to step 201 to select the next semantic relation (SR m ) of the morpheme, else the function continues at step 206.
- step 206 if the selected semantic relation of the derived word (SR d ) contains a semantic record (REC d ) with a lemma value (LV d ) equal to a lemma value (LV m ) of the selected semantic relation (SR m ) of the morpheme, then the function continues at step 207, else the function continues at step 208.
- step 207 the function increments the derivational score by the Morels value of the selected semantic relation (SR m ) of the morpheme and continues at step 212.
- step 208 if the selected semantic relation (SR m ) of the morpheme is a type equal to "Hypernym”, then the function continues at step 209, else the function continues at step 210.
- the system ensures that the derived word and base word combined with the morpheme are species of the same genus. Thus, when lemma values for "Hypernym" semantic relations are not equal, the system determines whether the lemma value for the derived word is related to the base word.
- the system determines this relatedness by comparing the "Hypernym" lemma values of the head word entry for the lemma value of the derived word to the lemma value of the morpheme.
- One skilled in the art would appreciate that other techniques may be used for checking whether the putative base word, in combination with the morpheme, is within the same genus as the derived word.
- the system could be adapted to check "Synonym" semantic relations.
- the function retrieves the dictionary entry for the head word equal to the lemma value (LV d ) of the selected semantic record (SR d ) .
- the function increments the derivational score by the Morels value of the selected semantic record (REC m ) and continues at step 212.
- the lemma value (LV m ) of the semantic record (REC m ) indicates the base word, then the function continues at step 211, else the function continues at step 212.
- the function determines whether the lemma value (LV d ) is related to the base word.
- This relatedness of words can be determined in several ways.
- a derived word is related to a base word when the base word contains a semantic relation type (regardless of nesting) equal to the semantic relation type of the selected semantic relation (SR m ) of the morpheme and the lemma value (LV d ) of the derived word equals the lemma value (LV b ) of the base word. If the base word is related to the derived word, then the function increments the derivational score by the Morels value in the selected semantic record (REC m ) of the morpheme.
- the function invokes the function CompareContent recursively passing the list of nested semantic relations (SR m ) of the morpheme and the list of nested semantic relations (SR d ) of the derived word.
- the system determines whether a sense of a derived word is derived from a sense of putative base word and the "er" morpheme.
- These examples illustrate (1)when the semantic relations in the derived word and those in the morpheme are structurally isomorphic and the lemma values in the semantic relations of the derived word are equivalent to those in the morpheme, (2)when the "Hypernym” semantic relations in the derived word and the morpheme have non-equal but related values, (3) when the "non-Hypernym” semantic relations in the derived word and the morpheme have non-equal but related values, and (4) when there is no relatedness.
- the system calculates a the word derivational score for the word "geographer” analyzed as the base word “geography” combined with the "er” morpheme.
- Table 4 contains the entry for the "er” morpheme with the base word “geography” inserted, and entry for the word "geographer”. To determine the score, the system applies the function CompareContent to a list of the semantic relations of the morpheme and a list of semantic relations of "geographer.” The system first selects the "Hypernym” semantic relation of the morpheme and the “Hypernym” semantic relation of "geographer.” The following chart shows this selection.
- the system determines that the lemma values "person" are equal, and increases the derivational score by the Morels value of 2 for a total of 2.
- the system processes each semantic record of "geographer."
- the following chart shows the first semantic relation.
- the derivational score is not increased.
- the system determines if the nested semantic relations of the morpheme matcha nested semantic relation of "geographer.” Since both the morpheme and “geographer” have a "HasObj" semantic relation with a lemma value of "geography,” the system increases the derivational score by the Morels value of 10 for a total of 12.
- the system then selects the second semantic record of "geographer."
- the following chart shows this selection.
- the derivational score is increased by the Morels value of 2 for a total of 14.
- the system determines if the nested semantic relations of the morpheme matcha nested semantic relation of "geographer.” Since both the morpheme and the entry for "geographer” have a "HasObj" semantic relation with a lemma value of "geography,” the system increases the derivational score by the Morels value of 10 for a total of 24. The derivational score of 24 indicates the likelihood that "geographer” is derived from "geography.”
- the system calculates a derivational score for the derived word "cartographer” analyzed as the base word “cartography” with the "er” morpheme.
- Table 5 contains the entry for the "er” morpheme with the base word “cartography” inserted, and the entry for "cartographer".
- the system applies the function CompareContent to a list of the semantic relations of the morpheme and a list of semantic relations of "cartographer.” The system first selects the "Hypernym” semantic relation of the morpheme and the “Hypernym” semantic relation of "cartographer.” The following chart shows this selection.
- the system determines that the lemma values "person" are equal, and increases the derivational score by the Morels value of 2 for a total of 2.
- the system processes each semantic record of "cartographer.” Since the lemma value "make” is not in the list of lemma values of the morpheme, the derivational score is not increased. The system then determines if the nested semantic relations of the morpheme matches a nested semantic relations of "cartographer.” Since both the morpheme and “cartographer” have an "HasObj" semantic relation, the system determines that the lemma value of "cartography” is not equal to the lemma value of "map.” The system then determines whether the base word "cartography" is related to the lemma value "map” by retrieving the entry for cartography and examining its semantic relation information. The following chart show the dictionary entry for
- the system determines that "cartography” has a "HasObj" (nested) semantic relation with a lemma value equal to "map.” The system then increments the derivational score by the Morels value of 10 for a total of 12. The derivational score of 12 indicates the likelihood that "cartographer” is derived from "cartography.”
- the system calculates a derivational score for the derived word "banker” analyzed as the base word “bank” with the "er” morpheme.
- This example uses the following sense of "banker:” "the player who keeps the bank in various games of
- Table 6 contains the entry for the "er” morpheme with the base word “bank” inserted and entry for the word "banker”. To determine the score, the system applies the function CompareContent to the list of semantic relations of the morpheme and the list of semantic relations of "banker.” The system first selects the "Hypernym” semantic relation of the morpheme and the “Hypernym” semantic relation of "banker.” The following chart shows this selection.
- the system determines that the lemma value "person” is not equal to the lemma value "player.” Because there is a mismatch on the "Hypernym” semantic relation, the system then retrieves the dictionary entry for "player.” The following shows the dictionary entry for
- the system processes each semantic record of "banker.” Since the lemma value "keep” is not in the list of lemma values of the morpheme, the derivational score is not increased. The system then determines if the nested semantic relations of the morpheme matches a nested semantic relation of "banker.” Since both the morpheme and "banker” have a "HasObj" semantic relation with a lemma value of "bank,” the system increases the derivational score by the Morels value of 10 for a total of 12.
- the system calculates a derivational score the derived word "comer” and the base word “corn” with the "er” morpheme.
- This example uses the following sense of "corner:” "the place where 2 roads, paths, or streets
- Table 7 contains the entry for the "er” morpheme with the base word “corn” inserted, and the entry for the word "corner”.
- the system applies the function CompareContent to a list of the semantic relations of the morpheme and a list of semantic relations of "corner.” The system first selects the "Hypernym” semantic relation of the morpheme and the “Hypernym” semantic relation of "corner.” The following chart shows this selection.
- the system determines that the lemma value "person” is not equal to the lemma value "place.” Because there is a mismatch on the "Hypernym” semantic relation, the system then retrieves the dictionary entry for "place.” Although not shown, the dictionary entry for "place” does not have a "Hypernym” semantic relation with a lemma value of "person.” Therefore, the system does not increase the derivational score.
- the system selects the "SubjOf" semantic relation of the morpheme. Since "corner” has no "SubjOf” semantic relation, the system does not increase the derivational score.
- the system adds a negative value to the derivational score when the derived word has no corresponding semantic relation of the same type as a semantic relation of the morpheme. The adding of a negative value to the derivational score indicates that the likelihood of the derivation is reduced.
- Table 8 contains pseudo-code for a more detailed algorithm used by the system to generate a derivational score indicating the likelihood that a sense of a derived word (D) can be correctly analyzed as a morpheme (M) combined with a sense of putative base word (B).
- the function CompareContent recursively calls itself to process nested semantic relations.
- the function IsRelated also recursively calls itself to determine whether the base word is related to the derived word. As shown, the function IsRelated checks for a semantic relation at any nesting level that has a type equal to the type of SRd and with a lemma value equal to the passed lemma value.
- the function IsRelated locates a structurally isomorphic position within the base word and checks for a related semantic relation. Also, in a preferred embodiment, each semantic relation of a morpheme contains an indicator of the semantic relation type that is structurally relevant. The system then checks for structural isomorphism using the indicator.
- the database is updated to include links from derived words to base words.
- the system stores the derivational score for each sense of derived words and base words.
- the system adjusts the derivational score based on (1) syntactic relatedness between the derived word and the base word, for example, by matching subcategorized prepositions on a verb sense and its derived nominalization and (2) matching other attributes between the base word and derived word.
- the scope of the present invention is defined by the claims that follow.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Machine Translation (AREA)
Abstract
A method and system for determining the derivational relatedness of a derived word and a base word. In a preferred embodiment, the system includes a machine-readable dictionary containing entries for head words and morphemes. Each entry contains definitional information and semantic relations. Each semantic relation specifies a relation between the head word with a word used in its definition. Semantic relations may contain nested semantic relations to specify relations between words in the definition. The system compares the semantic relations of the derived word to the semantic relations of a morpheme, which is putatively combined with the base word when forming the derived word. The system then generates a derivational score that indicates the confidence that the derived word derives from the base word.
Description
This invention relates generally to a computer system for deriving semantic information from machine-readable dictionaries, and more specifically, to a method and system for identifying morphologically related words in machine-readable dictionaries.
Machine-readable dictionaries (MRDs) often contain definitional information for each word in such a way that that information is accessible by a computer program. Typically, MRDs contain explicit and complete information making it possible to identify inflected forms of words. That information may appear in the form of orthographic rules or as a list of regular and irregular forms. Regular inflectional forms can be derived from rules in the "Guide to the Dictionary," which is typically included with the dictionary. Irregular inflectional forms are explicitly listed in dictionary entries. For example, the entry for the verb "sing" lists the irregular forms "sang" and "sung" and the regular forms "sings" and "singing."
However, MRDs typically contain only implicit and scant information or word formation (derivation). Thus, it is difficult for even a human user to construct a list of all words derived from a given word. For example, the following words, which are morphologically related to the verb "believe," are defined separately in typical MRDs: "belief," "believe," "believable," "unbelievable," "believer," "disbelief," "disbelieve," "unbelief," "unbelieving," and "unbelievingly." These words are not marked explicitly in the dictionary as being morphologically related. The word "believe" is the base word of these derived words.
The word "believer" is formed from the base word "believe" and the morpheme "er." A morpheme is a meaningful linguistic unit that cannot be further divided into meaningful parts. Automated methods have been developed for parsing a derived word into its component parts: putative base word and morpheme. Generally, these methods use finite state morphological analysis as described in Sproat, R., "Morphology and Computation," The MIT Press, 1992. Finite state morphological analysis applies morpheme-based rules to identify the component parts. For example, one rule may be that the morpheme "ize" attached to a noun or an adjective creates a verb. Thus, the word "hospitalize" can be parsed into "hospital" and "ize." Conversely, the word "maize" would not parse into "ma" and "ize," because "maize" is a noun and not a verb. However, successful parsing of a word does not guarantee a morphological relationship between the word and its putative base word. For example, the word "flower" may parse into "flow" and "er," but the word "flower" is not morphologically related to the word "flow."
MRDs, however, may include some form of overt markings of morphological relatedness. Specifically, some undefined forms may be listed at the end of the dictionary entry for the word from which they are ultimately derived. The MRDs assume that the meaning of these words should be clear when the meaning of the suffix is added to the meaning of the base word. For example, the dictionary entry for the word "journalism" may indicate that the form "journalistic" is derived from that word.
These overt markings, however, do not provide complete or systematic list of morphologically related words. Natural language processing systems, which need to model the knowledge of the speakers of the language, can benefit from access to complete and systematic morphological information. Although this information can be derived manually, the size of MRDs makes this derivation time-consuming and makes it difficult to ensure completeness. It would be desirable to have an automated method for deriving morphological information from a MRD. Further, it would be desirable to have an automated method for determining when a word does not derive from a putative base word. For example, this method would automatically determine that the noun "flower" does not derive from the verb "flow."
The present invention provides a method and system for determining the derivational relatedness of a derived word and a base word. In a preferred embodiment, the system includes a database with an entry for each head word and morpheme. Each entry contains definitional information from a machine-readable dictionary and semantic relations extracted from the definitions. Each semantic relation specifies a relation between the head word with a word used in its definition. Semantic relations may contain nested semantic relations to specify relations between words in the definition. The system identifies a putative base word of a derived word preferably using finite state morphological analysis. The system then compares the semantic relations of the derived word to the semantic relations of the morpheme, which is augmented with the base word. The system generates a derivational score that indicates the confidence that the derived word derive from the base word.
FIG. 1 is a block diagram of a computer system of a preferred embodiment.
FIG. 2 is a flow diagram of a function CompareContent.
The present invention provides a system and method for automatically identifying morphologically related words. In a preferred embodiment, the system, known as Morels for morphological relation system, determines whether a word, known as a derived word, is indeed derived from a putative base word identified preferably using finite state morphological analysis. The system assigns a derivational score to the putative derivation indicating the likelihood that the derived word is a derivative of the base word. To determine the score, the system preferably uses a database containing semantic relations for each head word and morpheme in a machine-readable dictionary (MRD). A head word is a word for which the dictionary contains an entry. A morpheme is a meaningful linguistic unit that cannot be divided into further meaningful parts (e.g., "er," "ed," "ing," and "ion"). A derived word is considered to be a combination of the putative base word and a morpheme. The system compares the semantic relations of the morpheme augmented with the base word to the semantic relations for the derived word. A semantic relation (e.g., genus, subject, object) expresses a relation between a head word and a word in the definition. In a preferred embodiment, the semantic relations are generated automatically by parsing the definition of a head word. Based on the comparison, the system assigns a derivational score to the putative derivation. This derivational score indicates the likelihood that the derived word derives from the putative base word.
FIG. 1 is a block diagram of a computer system of a preferred embodiment. The computer system includes a central processing unit 101, a memory 102, a database 103, and a display 104. The memory contains executable computer code 105 implementing a preferred embodiment of the present invention. The database contains a dictionary with semantic relations specified.
In a preferred embodiment, the database contains an entry for each head word and morpheme of a dictionary. Each head word entry contains a list of semantic relations for each sense (different definition) of the head word. Each semantic relation comprises an attribute (type) and has as its value a list of one or more semantic records. A semantic record contains a lemma value and optionally one or more nested semantic relations. A lemma value is the uninflected form of the word, which is typically the form under which the word is listed in the dictionary as a head word. For example, the lemma of "takes," "taking," "took," and "taken" is the form "take" and the lemma of "children" is
TABLE 1
______________________________________
{HeadWord "banker"
Senses
Defin "a person who owns or controls or shares in the control of a
bank"
Hypernym
{Lemma "person"}
SubjOf
{Lemma "own"
HasObj
{Lemma "bank"}}
{Lemma "control"
HasObj
{Lemma "bank}}
{Lemma "share"
HasObj
{Lemma "bank}}
{Defin "the player who keeps the bank in various games of chance"
Hypernym
{Lemma "player"}
SubjOf
{Lemma "keep"
LocatedAt
{Lemma "game"
PartOf
{Lemma "chance"}}
HasObj
{Lemma "bank"}}
}
}
______________________________________
Table 1 contains the entry for the head word "banker." The entry contains two definitions or senses: (1) "a person who owns or controls or shares in the control of a bank" and (2) "the player who keeps the bank in various games of chance." Each sense contains the definition of the sense and the list of semantic relations for that sense. The first sense contains two semantic relations. The first semantic relation is
______________________________________
Hypernym
{Lemma "person"}
______________________________________
The semantic relation type is "Hypernym," and its value is the semantic record "{Lemma "person"}." The lemma value is "person," and the semantic record contains no nested semantic relations. The second semantic relation is
______________________________________
SubjOf
{Lemma "own"
HasObj
{Lemma "bank"}}
{Lemma "control"
HasObj
{Lemma "bank}}
{Lemma "share"
HasObj
{Lemma "bank}}
______________________________________
The semantic relation type is "SubjOf," and its value is a list of three semantic records. Each semantic record contains a nested semantic relation with the semantic relation type of "HasObj."
An unnested semantic relation relates to the head word (the related-to word). A nested semantic relation relates to the lemma value (the related-to word) of the nesting semantic record. A "Hypernym" type identifies that the lemma value is the genus for which the related to word is a species. A "SubjOf" type identifies that the related-to word is the subject of the lemma value. A "HasObj" type identifies that the lemma value is the object of the related-to word. A "PartOf" type identifies that the related-to word is part of the lemma value. For example, the semantic relations for the first sense of "banker" indicate that "banker" is a kind of "person" (Hypernym), that "banker" is the subject of "own" (SubjOf, and that "bank" is the object of "own" (HasObj). Table 2 contains examples illustrating the meaning of various types of semantic relations.
TABLE 2
______________________________________
SEMANTIC
RELATIONS EXAMPLE
301 EXAMPLE DEFINITION
______________________________________
Actor baton --(Actor)→
baton: a short thick stick
policeman used as a weapon by a
policeman
ByMeansOf
boil --(ByMeansOf)→
to boil: to cause to reach
cook the stated condition by
cooking in water
language --(ByMeansOf)→
language: the system of
word human expression by
means of words
CausedBy disease --(CausedBy)→
disease: (an) illness or
infection, growth
disorder caused by in-
fection or unnatural
growth, not by an
accident
Causes germ --(Causes)→
germ: a very small living
disease thing which cannot be
seen but may live on
food or dirt or in the
body, so causing disease
DegreeOf height --(DegreeOf)→
height: the quality or
tall, high degree of being tall or
high
Domain stock --(Domain)→
stock: a liquid made from
cook the juices of meat, bones,
etc. used in cooking
HasObj reef (HasObj)→ sail
to reef: to tie up (part of
a sail) so as to reduce
the size
HasPart oak --(HasPart)→ wood
oak: any of several types
of large tree with hard
wood
HasSubj gallop --(HasSubj)→
to gallop: (of a horse) to
horse go at the fastest speed
Hypernym oak --(Hypernym)→ tree
oak: any of several types
of large tree with hard
wood
InstrumentFor
hammer --(InstrumentFor)→
hammer: a tool with a
drive, strike heavy head for driving
nails into wood, or for
striking things to break
them or move them
IsFor facing --(IsFor)→
facing: an outer covering
protection, ornament
or surface, as of a wall,
for protection, ornament,
etc.
fair --(IsFor)→ sell
fair: a market, esp. one
held at a particular place
at regular periods for
selling farm produce
LocatedAt
hammer --(LocatedAt)→
hammer: a bone in the
ear ear
LocationOf
market --(LocationOf)→
market: a building,
meet square, or open place
where people meet to
buy and sell goods, esp.
food, or sometimes
animals
MadeInto nylon --(MadeInto)→
nylon: a strong elastic
cloth, cord, plastic
material made from coal,
water, and air and made
into cloth, cords,
plastics
MadeOf nylon --(MadeOf)→
nylon: a strong elastic
coal, water, air material made from coal,
water, and air and made
into cloth, cords,
plastics
Manner gawp --(Manner)→
to gawp: to look at some-
foolish thing in a foolish way,
esp. with the mouth open
pleasure --(Manner)→
pleasure: the state or
happiness, satisfaction
feeling of happiness or
satisfaction
ObjOf arbitrator --(ObjOf)→
arbitrator: a person
choose chosen by both sides of
an argument to examine
the facts and make a
decision to settle the
argument
PartOf bone --(PartOf)→
bone: any of numerous
skeleton anatomically distinct
structures making up the
skeleton of a vertebrate
animal
Possessor
navy --(Possessor)→
navy: the ships of war
country belonging to a country
Purpose additive --(Purpose)→
additive: a substance
improve, add added in small quantities
to something else, as to
improve the quality, or
add color, taste, etc.
SubjOf riveter --(SubjOf)→
riveter: a person whose
fasten job is fastening rivets
fruit --(SubjOf)→
fruit: an object that
grow, contain grows on a tree or a bush
and contains seeds
Synonym disposal--(Synonym)→
disposal: arrangement
arrangement
lyricist --(Synonym)→
lyricist: a writer of lyrics;
songwriter songwriter
TimeOf summer --(TimeOf)→
summer: the season
hot, flower between spring and
autumn when the sun is
hot and there are many
flowers
______________________________________
The semantic relations for morphemes are similar to those for head words. Table 3 contains the morpheme entry for the morpheme
TABLE 3
______________________________________
{Morpheme "er"
{Defin "a person who knows about or works at"
Hypernym
{Lemma "person"
Morels 2}
SubjOf
{Lemma ("know", "work")
Morels 2
HasObj
{Lemma "base word"
Morels 10}}}
______________________________________
Morpheme semantic relations contain Morels values. A Morels value indicates the amount by which a derivational score is to be increased when structural isomorphism and equivalent lemma values are encountered during the evaluation of morphological relatedness between the derived word and the putative base word. Structural isomorphism occurs when words have similar hierarchy of semantic relation types. For example, a head word entry with a "Hypernym" and "SubjOf" semantic relation and with a "HasObj" semantic relation nested within the "SubjOf" semantic relation is structurally isomorphic to the morpheme "er." The lemma value of "base word" indicates that the putative base word should be inserted to complete the semantic relations for the putative base word combined with the morpheme. As shown in Table 3, the lemma value for a morpheme may be represented by a list of lemma values ("know", "work") when their semantic records are structurally isomorphic with equivalent lemma values.
The semantic relations for each head word of dictionary database may be generated manually or automatically. In a preferred embodiment, the system uses a database generated as described by Dolan, Vanderwende, and Richardson in Proceedings of the First Conference of the Pacific Association for Computational Linguistics, 1993, pp. 5-14; Jensen and Binot, "A Semantic Expert Using On-Line Standard Dictionary," reprinted in Natural Language Processing: The PLNLP Approach, 1992, chap. 11; and Montemagni and Vanderwende, Proceedings of Coling, 1992, pp. 546-52, all of which are hereby incorporated by reference. Since the number of morphemes in a dictionary is relatively small, the Morels values and the morpheme semantic relations are preferably manually entered into the database. Although the preferred embodiment is described using a MRD with semantic relations already specified, one skilled in the art would appreciate that the semantic relation information can be dynamically generated during derivational analysis from definitional information in the MRD.
The system generates a derivational score for a sense of a derived word that indicates the likelihood that the derived word is a derivation of a putative base word and a morpheme. In a preferred embodiment, the system generates an instance of the morpheme semantic relations inserting the putative base word where indicated. The system then selects each semantic relation of the instantiated morpheme and determines whether the derived word has a semantic relation of the same type. If the derived word has such a semantic relation and the lemma value of the instantiated morpheme and the derived word are equal, then the system increases the derivational score. If the lemma values are not equal and the semantic relation type is a "Hypernym," then the system determines whether the lemma value of the derived word is related to the lemma value of the instantiated morpheme. If the semantic relation type is not "Hypernym," then the system determines if the base word is related to the lemma value of the derived word. The system then recursively determines whether nested semantic relations of the instantiated morpheme are related to the nested semantic relations of the derived word.
FIG. 2 is a flow diagram illustrating a CompareContent function of the system. The function is passed a list of semantic relations for a morpheme (LSRm) and a list of semantic relations for a derived word (LSRd). The function determines derivational relatedness of the derived word and the putative base word. The function recursively invokes itself to process nested semantic relations. The function is initially invoked with a global derivational score set to zero. The function increases the derivational score to indicate derivational relatedness. In step 201, the function selects the next semantic relation (SRm) of the morpheme starting with the first semantic relation in LSRm. The function loops processing each semantic relation in the morpheme. In step 202, if all the semantic relations for the morpheme have already been selected, then the function returns, else the function continues at step 203. In step 203, if the list of semantic relations (LSRd) for the derived word contains a semantic relation (SRd) with a type equal to the selected semantic relation (SRm) of the morpheme, then the function continues at step 204, else the function loops to step 201 to select the next semantic relation (SRm) of the morpheme. In step 204, the function selects the next semantic record (RECd) of the selected semantic relation (SRd) of the derived word starting with the first semantic record. In step 205, if all semantic records (RECd) have already been selected, then the function loops to step 201 to select the next semantic relation (SRm) of the morpheme, else the function continues at step 206. In step 206, if the selected semantic relation of the derived word (SRd) contains a semantic record (RECd) with a lemma value (LVd) equal to a lemma value (LVm) of the selected semantic relation (SRm) of the morpheme, then the function continues at step 207, else the function continues at step 208. In step 207, the function increments the derivational score by the Morels value of the selected semantic relation (SRm) of the morpheme and continues at step 212. In step 208, if the selected semantic relation (SRm) of the morpheme is a type equal to "Hypernym", then the function continues at step 209, else the function continues at step 210. In a preferred embodiment, the system ensures that the derived word and base word combined with the morpheme are species of the same genus. Thus, when lemma values for "Hypernym" semantic relations are not equal, the system determines whether the lemma value for the derived word is related to the base word. The system determines this relatedness by comparing the "Hypernym" lemma values of the head word entry for the lemma value of the derived word to the lemma value of the morpheme. One skilled in the art would appreciate that other techniques may be used for checking whether the putative base word, in combination with the morpheme, is within the same genus as the derived word. For example, the system could be adapted to check "Synonym" semantic relations. In step 209, the function retrieves the dictionary entry for the head word equal to the lemma value (LVd) of the selected semantic record (SRd) . If the retrieved dictionary entry contains a semantic relation (SRh) with a type of "Hypernym," and the semantic relation (SRh) contains a semantic record (RECh) with its lemma value (LVh) equal to the lemma value (LVm) of the morpheme, then the function increments the derivational score by the Morels value of the selected semantic record (RECm) and continues at step 212. In step 210, if the lemma value (LVm) of the semantic record (RECm) indicates the base word, then the function continues at step 211, else the function continues at step 212. In step 211, the function determines whether the lemma value (LVd) is related to the base word. This relatedness of words can be determined in several ways. In a preferred embodiment, a derived word is related to a base word when the base word contains a semantic relation type (regardless of nesting) equal to the semantic relation type of the selected semantic relation (SRm) of the morpheme and the lemma value (LVd) of the derived word equals the lemma value (LVb) of the base word. If the base word is related to the derived word, then the function increments the derivational score by the Morels value in the selected semantic record (RECm) of the morpheme. In step 212, the function invokes the function CompareContent recursively passing the list of nested semantic relations (SRm) of the morpheme and the list of nested semantic relations (SRd) of the derived word.
In the following examples, the system determines whether a sense of a derived word is derived from a sense of putative base word and the "er" morpheme. These examples illustrate (1)when the semantic relations in the derived word and those in the morpheme are structurally isomorphic and the lemma values in the semantic relations of the derived word are equivalent to those in the morpheme, (2)when the "Hypernym" semantic relations in the derived word and the morpheme have non-equal but related values, (3) when the "non-Hypernym" semantic relations in the derived word and the morpheme have non-equal but related values, and (4) when there is no relatedness. In the first example in which the semantic relations of the derived word are structurally isomorphic to those in the morpheme and the lemma values in the semantic relations of the two are equivalent, the system calculates a the word derivational score for the word "geographer" analyzed as the base word "geography" combined with the "er" morpheme.
TABLE 4
______________________________________
{Morpheme "er"
{Defin "a person who knows about or works at"
Hypernym
{Lemma "person"
Morels 2}
SubjOf
{Lemma ("know", "work")
Morels 2
HasObj
{Lemma "geography"
Morels 10}}}
{HeadWord "geographer"
Senses
{Defin "a person who studies and knows about geography"
Hypernym
{Lemma "person"}
SubjOf
{Lemma "study"
HasObj
{Lemma "geography"}}
{Lemma know"
HasObj
{Lemma "geography"}}}
}
______________________________________
Table 4 contains the entry for the "er" morpheme with the base word "geography" inserted, and entry for the word "geographer". To determine the score, the system applies the function CompareContent to a list of the semantic relations of the morpheme and a list of semantic relations of "geographer." The system first selects the "Hypernym" semantic relation of the morpheme and the "Hypernym" semantic relation of "geographer." The following chart shows this selection.
______________________________________
"er" "geographer"
______________________________________
Hypernym Hypernym
{Lemma "person" {Lemma "person"}
Morels 2}
______________________________________
The system then determines that the lemma values "person" are equal, and increases the derivational score by the Morels value of 2 for a total of 2.
The system then selects the "SubjOf" semantic relation in the morpheme and the "SubjOf" semantic relation of "geographer." The following chart shows this selection.
______________________________________
"er" "geographer"
______________________________________
SubjOf SubjOf
{Lemma ("know", "work")
{Lemma "study"
Morels 2
HasObj HasObj
{Lemma "geography" {Lemma "geography"}}
Morels 10}}
{Lemma "know"
HasObj
{Lemma "geography"}}
______________________________________
The system processes each semantic record of "geographer." The following chart shows the first semantic relation.
______________________________________
"er" "geographer"
______________________________________
SubjOf SubjOf
{Lemma ("know", "work")
{Lemma "study"
Morels 2
HasObj HasObj
{Lemma "geography" {Lemma "geography"}}
Morels 10}}
______________________________________
Since the lemma value "study" is not in the list of lemma values of the morpheme, the derivational score is not increased. The system then determines if the nested semantic relations of the morpheme matcha nested semantic relation of "geographer." Since both the morpheme and "geographer" have a "HasObj" semantic relation with a lemma value of "geography," the system increases the derivational score by the Morels value of 10 for a total of 12.
The system then selects the second semantic record of "geographer." The following chart shows this selection.
______________________________________
"er" "geographer"
______________________________________
SubjOf SubjOf
{Lemma ("know", "work")
{Lemma "know"
Morels 2
HasObj HasObj
{Lemma "geography" {Lemma "geography"}}
Morels 10}}}
______________________________________
Since the lemma value "know" is in the list of lemma values of the morpheme, the derivational score is increased by the Morels value of 2 for a total of 14.
The system then determines if the nested semantic relations of the morpheme matcha nested semantic relation of "geographer." Since both the morpheme and the entry for "geographer" have a "HasObj" semantic relation with a lemma value of "geography," the system increases the derivational score by the Morels value of 10 for a total of 24. The derivational score of 24 indicates the likelihood that "geographer" is derived from "geography."
In the second example in which "non-Hypernym" semantic relations have non-equal but related lemma values, the system calculates a derivational score for the derived word "cartographer" analyzed as the base word "cartography" with the "er" morpheme.
TABLE 5
______________________________________
{Morpheme "er"
{Defin "a person who knows about or works at"
Hypernym
{Lemma "person"
Morels 2}
SubjOf
{Lemma ("know", "work")
Morels 2
HasObj
{Lemma "cartography"
Morels 10}}}
{HeadWord "cartographer"
Senses
{Frgom "a "person who makes maps"
Hypernym
Lemma "person"}
SubjOf
{Lemma "make"
HasObj
{Lemma "map"}}}
}
______________________________________
Table 5 contains the entry for the "er" morpheme with the base word "cartography" inserted, and the entry for "cartographer". To determine the score, the system applies the function CompareContent to a list of the semantic relations of the morpheme and a list of semantic relations of "cartographer." The system first selects the "Hypernym" semantic relation of the morpheme and the "Hypernym" semantic relation of "cartographer." The following chart shows this selection.
______________________________________
"er" "cartographer"
______________________________________
Hypernym Hypernym
{Lemma "person" Lemma "person"}
Morels 2}
______________________________________
The system then determines that the lemma values "person" are equal, and increases the derivational score by the Morels value of 2 for a total of 2.
The system then selects the "SubjOf" semantic relation of the morpheme and the "SubjOf" semantic relation of "cartographer." The following chart shows this selection.
______________________________________
"er" "cartographer"
______________________________________
SubjOf SubjOf
{Lemma ("know", "work")
{Lemma "make"
Morels 2
HasObj HasObj
{Lemma "cartography" {Lemma "map"}}
Morels 10}}
______________________________________
The system processes each semantic record of "cartographer." Since the lemma value "make" is not in the list of lemma values of the morpheme, the derivational score is not increased. The system then determines if the nested semantic relations of the morpheme matches a nested semantic relations of "cartographer." Since both the morpheme and "cartographer" have an "HasObj" semantic relation, the system determines that the lemma value of "cartography" is not equal to the lemma value of "map." The system then determines whether the base word "cartography" is related to the lemma value "map" by retrieving the entry for cartography and examining its semantic relation information. The following chart show the dictionary entry for
______________________________________
{HeadWord "cartography"
Senses
{Defin "the science or art of making maps"
Hypernym
{Lemma "make"
Classifier
{Lemma "art"}
{Lemma "science"}
HasObj
{Lemma "map"}
______________________________________
The system determines that "cartography" has a "HasObj" (nested) semantic relation with a lemma value equal to "map." The system then increments the derivational score by the Morels value of 10 for a total of 12. The derivational score of 12 indicates the likelihood that "cartographer" is derived from "cartography."
In the third example in which "Hypernym" semantic relations have non-equal but related values, the system calculates a derivational score for the derived word "banker" analyzed as the base word "bank" with the "er" morpheme. This example uses the following sense of "banker:" "the player who keeps the bank in various games of
TABLE 6
______________________________________
{Morpheme "er"
{Defin "a person who knows about or works at"
Hypernym
{Lemma "person"
Morels 2}
SubjOf
{Lemma ("know", "work")
Morels 2
HasObj
{Lemma "bank"
Morels 10}}}
{HeadWord "banker"
Senses
{Defin "the player who keeps the bank in various games of chance"
Hypernym
{Lemma "player"}
SubjOf
{Lemma "keep"
LocatedAt
{Lemma "game"
PartOf
{Lemma "chance"}}
HasObj
{Lemma "bank"}}}
}
______________________________________
Table 6 contains the entry for the "er" morpheme with the base word "bank" inserted and entry for the word "banker". To determine the score, the system applies the function CompareContent to the list of semantic relations of the morpheme and the list of semantic relations of "banker." The system first selects the "Hypernym" semantic relation of the morpheme and the "Hypernym" semantic relation of "banker." The following chart shows this selection.
______________________________________
"er" "bank"
______________________________________
Hypernym Hypernym
{Lemma "person" {Lemma "player"}
Morels 2}
______________________________________
The system then determines that the lemma value "person" is not equal to the lemma value "player." Because there is a mismatch on the "Hypernym" semantic relation, the system then retrieves the dictionary entry for "player." The following shows the dictionary entry for
______________________________________
{HeadWord "player"
Senses
{Defin "a person taking part in a game or sport"
Hypernym
{Lemma "person"}
SubjOf
{Lemma "take"
HasObj
{Lemma "part"
LocatedAt
{Lemma "game"}
{Lemma "sport"}}}
}
______________________________________
Since the "Hypernym" semantic relation for "player" contains a lemma value of "person," which is equal to the lemma value for the morpheme, the system increases the derivational score by the Morels value of 2 for a total of 2.
The system then selects the "SubjOf" semantic relation of the morpheme and the "SubjOf" semantic relation of "banker." The following chart shows this selection.
______________________________________
"er" "banker"
______________________________________
SubjOf SubjOf
{Lemma ("know", "work")
{Lemma "keep"
Morels 2 LocatedAt
{Lemma "game"
PartOf
{Lemma "chance"}}
HasObj HasObj
{Lemma "bank" {Lemma "bank"}}
Morels 10}}
______________________________________
The system processes each semantic record of "banker." Since the lemma value "keep" is not in the list of lemma values of the morpheme, the derivational score is not increased. The system then determines if the nested semantic relations of the morpheme matches a nested semantic relation of "banker." Since both the morpheme and "banker" have a "HasObj" semantic relation with a lemma value of "bank," the system increases the derivational score by the Morels value of 10 for a total of 12.
In the fourth example in which there is no relatedness, the system calculates a derivational score the derived word "comer" and the base word "corn" with the "er" morpheme. This example uses the following sense of "corner:" "the place where 2 roads, paths, or streets
TABLE 7
______________________________________
{Morpheme "er"
{Defin "a person who knows about or works at"
Hypernym
{Lemma "person"
Morels 2}
SubjOf
{Lemma ("know", "work")
Morels 2
HasObj
{Lemma "corn"
Morels 10}}}
{HeadWord corner
{Defin "the place where 2 roads, paths, or streets meet"
Hypernym
{Lemma "place"
LocationOf
{Lemma "meet"
HasSubj
{Lemma "road"}
{Lemma "path"}
{Lemma "street"}}}
}
______________________________________
Table 7 contains the entry for the "er" morpheme with the base word "corn" inserted, and the entry for the word "corner". To determine the score, the system applies the function CompareContent to a list of the semantic relations of the morpheme and a list of semantic relations of "corner." The system first selects the "Hypernym" semantic relation of the morpheme and the "Hypernym" semantic relation of "corner." The following chart shows this selection.
______________________________________
"er" "corner"
______________________________________
Hypernym
{Lemma "person" Hypernym
Morels 2} {Lemma "place"
______________________________________
The system then determines that the lemma value "person" is not equal to the lemma value "place." Because there is a mismatch on the "Hypernym" semantic relation, the system then retrieves the dictionary entry for "place." Although not shown, the dictionary entry for "place" does not have a "Hypernym" semantic relation with a lemma value of "person." Therefore, the system does not increase the derivational score.
The system then selects the "SubjOf" semantic relation of the morpheme. Since "corner" has no "SubjOf" semantic relation, the system does not increase the derivational score. In an alternate embodiment, the system adds a negative value to the derivational score when the derived word has no corresponding semantic relation of the same type as a semantic relation of the morpheme. The adding of a negative value to the derivational score indicates that the likelihood of the derivation is reduced.
Table 8 contains pseudo-code for a more detailed algorithm used by the system to generate a derivational score indicating the likelihood that a sense of a derived word (D) can be correctly analyzed as a morpheme (M) combined with a sense of putative base word (B). The function CompareContent recursively calls itself to process nested semantic relations. The function IsRelated also recursively calls itself to determine whether the base word is related to the derived word. As shown, the function IsRelated checks for a semantic relation at any nesting level that has a type equal to the type of SRd and with a lemma value equal to the passed lemma value. In an alternate embodiment, the function IsRelated locates a structurally isomorphic position within the base word and checks for a related semantic relation. Also, in a preferred embodiment, each semantic relation of a morpheme contains an indicator of the semantic relation type that is structurally relevant. The system then checks for structural isomorphism using the indicator.
TABLE 8
______________________________________
Definitions:
______________________________________
LSR.sub.d = list of semantic relations of D
SR.sub.d = semantic relation of LSR.sub.d
REC.sub.d = a record of semantic relation SR.sub.d
lemma.sub.-- value.sub.d = lemma value of REC.sub.d
instantiate M for Morpheme insert B in base word lemma.sub.-- values
Score = 0
CompareContent(LSR.sub.m, LSR.sub.d)
CompareContent(LSR.sub.m, LSR.sub.d)
{for each SR.sub.m of LSR.sub.m
if LSR.sub.d contains SR.sub.d equal to SR.sub.m
for each REC.sub.d of SR.sub.d
if RL.sub.m contains a lemma.sub.-- value.sub.m equals lemma.sub.--
value.sub.d
score += Morels
else if SR.sub.m is "Hypernym"
for each sense H for the head word of lemma.sub.-- value.sub.d
if LSR.sub.h contains SR.sub.h of "Hypernym"
for each REC.sub.h of SR.sub.h
if lemma.sub.-- value.sub.m equals lemma.sub.-- value.sub.h
1
score += Morels
endif
endfor
endif
endfor
else if lemma.sub.-- value.sub.m is base word
if (IsRelated(LSR.sub.b, SR.sub.d, lemma.sub.-- value.sub.d))
score += Morels
endif
endif
CompareContent(list of nested SR.sub.m, list of nested SR.sub.d)
endfor
endif
endfor
IsRelated(LSR.sub.b, SR.sub.d, lemma.sub.-- value.sub.d)
{for each SR.sub.b of LSR.sub.b
if SR.sub.b is equal to SR.sub.d
if lemma.sub.-- value.sub.b is equal to lemma.sub.-- value.sub.d
return (TRUE)
endif
endif
if (IsRelated(list of nested SR's of SR.sub.b, SR.sub.d, lemma.sub.--
value.sub.d))
return (TRUE)
endif
endfor
return (FALSE)
}
______________________________________
Although the present invention has been described in terms of preferred embodiments, it is not intended that the invention be limited to these embodiments. Modifications within the spirit of the invention will be apparent to those skilled in the art. For example, in a preferred embodiment, the database is updated to include links from derived words to base words. The system stores the derivational score for each sense of derived words and base words. In an alternate embodiment, the system adjusts the derivational score based on (1) syntactic relatedness between the derived word and the base word, for example, by matching subcategorized prepositions on a verb sense and its derived nominalization and (2) matching other attributes between the base word and derived word. The scope of the present invention is defined by the claims that follow.
Claims (28)
1. A method in a computer system for determining whether a derived word is morphologically related to a putative base word combined with a morpheme, the method comprising the steps of:
combining semantic information derived from a definition of the morpheme with the putative base word;
comparing semantic information derived from a definition of the derived word to the combined semantic information; and
determining whether the comparison indicates derivational relatedness of the derived word to the putative base word, the derivational relatedness indicating whether the derived word is morphologically related to the putative base word.
2. The method of claim 1 wherein the step of comparing includes the step of determining similarity of the semantic information for the derived word and the semantic information for the morpheme combined with the putative base word.
3. The method of claim 1 including the step of generating the semantic information for the derived word and the morpheme based on definitional information in a machine readable dictionary.
4. The method of claim 3 wherein the step of generating the semantic information from the definitional information is performed at the time that it is being determined whether the derived word is morphologically related to the putative base word combined with a morpheme.
5. The method of claim 3 wherein the step of generating the semantic information is performed during a preprocessing step in which the semantic information is then stored in the machine readable dictionary.
6. The method of claim 1 including the step of generating an indicator of how morphologically related the putative base word is to the derived word.
7. The method of claim 6 wherein the indicator indicates that the putative base word is not morphologically related to the derived word.
8. The method of claim 1 including the step of parsing the derived word to determine a putative base word.
9. The method of claim 1 including the step of determining a putative base word for the derived word using finite state morphological analysis.
10. The method of claims 3, 4, or 5 wherein the semantic information consists of semantic relations from definitional information.
11. A method in a computer system for determining whether a derived word is morphologically related to a putative base word combined with a morpheme, the method comprising the step of comparing semantic information for the derived word to semantic information for the morpheme combined with the putative base word wherein the comparison indicates derivational relatedness of the derived word to the putative base word, the derivational relatedness indicating whether the derived word is morphologically related to the putative base word wherein the semantic information for the derived word includes semantic relations, each semantic relation having a semantic relation type indicating a relationship between the derived word and a definitional word in definitional information for the derived word, and wherein the semantic information for the morpheme includes semantic relations, each semantic relation having a semantic relation type indicating a relationship between the morpheme and a definitional word in definitional information for the morpheme, and the step of comparing includes the steps of:
determining whether the derived word has semantic relations that are isomorphic to the semantic relations of the morpheme; and
when semantic relations are isomorphic, determining whether an uninflected form of the definitional word of the morpheme semantic relation is the same or is related to an uninflected form of the definitional word of the derived word semantic relation.
12. The method of claim 11 wherein the step of determining whether the morpheme semantic relation is related to the derived word semantic relation includes the step of, when the morpheme semantic relation and the derived word semantic relation have nested semantic relations, determining relatedness based on the nested semantic relations.
13. A method in a computer system for determining whether a derived word is morphologically related to a putative base word combined with a morpheme, the method comprising the step of comparing semantic information for the derived word to semantic information for the morpheme combined with the putative base word wherein the comparison indicates derivational relatedness of the derived word to the putative base word, the derivational relatedness indicating whether the derived word is morphologically related to the putative base word wherein the semantic information for the derived word includes semantic relations, each semantic relation having a semantic relation type indicating a relationship between the derived word and a definitional word in definitional information for the derived word, and wherein the semantic information for the morpheme includes semantic relations, each semantic relation having a semantic relation type indicating a relationship between the morpheme and a definitional word in definitional information for the morpheme, and the step of comparing includes the steps of:
determining whether the derived word has semantic relations that are isomorphic to the semantic relations of the morpheme; and
when semantic relations are isomorphic, determining whether the definitional word of the morpheme semantic relation is the same or is related to the definitional word of the derived word semantic relation such that when the semantic relation is a hypernym relation and the definitional words are not the same, determining relatedness based on the semantic information for the definitional word of the derived word semantic relation and the semantic information of the morpheme.
14. A method in a computer system for determining whether a derived word is morphologically related to a putative base word combined with a morpheme, the method comprising the step of comparing semantic information for the derived word to semantic information for the morpheme combined with the putative base word wherein the comparison indicates derivational relatedness of the derived word to the putative base word, the derivational relatedness indicating whether the derived word is morphologically related to the putative base word wherein the semantic information for the derived word includes semantic relations, each semantic relation having a semantic relation type indicating a relationship between the derived word and a definitional word in definitional information for the derived word, and wherein the semantic information for the morpheme includes semantic relations, each semantic relation having a semantic relation type indicating a relationship between the morpheme and a definitional word in definitional information for the morpheme, and the step of comparing includes the steps of:
determining whether the derived word has semantic relations that are isomorphic to the semantic relations of the morpheme; and
when semantic relations are isomorphic, determining whether the definitional word of the morpheme semantic relation is the same or is related to the definitional word of the derived word semantic relation such that when the semantic relation is a non-hypernym relation and the definitional words are not the same, determining relatedness based on the semantic information of the putative base word and the semantic information of the derived word.
15. The method of claim 11 or 14 including the step of using the uninflected form of the definitional word when determining whether two definitional words are the same or related.
16. A computer system for determining derivational relatedness among words, comprising:
a computer having a central processing unit and storage;
a dictionary, the dictionary being stored on the storage and having an entry for each head word of the dictionary, each entry having definitional information; and
a derivational computer program for receiving a derived word, a putative base word, and a morpheme, for combining semantic information derived from the definition of the morpheme with the putative base word, for comparing semantic information derived from the definition of the derived word with the combined semantic information and for generating, based on the comparison, an indicator of the derivational relatedness of the derived word to the putative base word, the derivational computer program executed in the computer system.
17. A computer system for determining derivational relatedness among words, comprising:
a computer having a central processing unit and storage;
a dictionary, the dictionary being stored on the storage and having an entry for each head word of the dictionary, each entry having definitional information; and
a derivational computer program for receiving a derived word, a putative base word, and a morpheme, for comparing semantic information derived from the definition of the derived word with semantic information derived from the definition of the morpheme combined with the putative base word, and for generating, based on the comparison, an indicator of the derivational relatedness of the derived word to the putative base word, the derivational computer program executed in the computer system wherein the derivational computer program stores the generated indicator in the dictionary to indicate that the derived word derives from the putative base word with an indicated degree of confidence.
18. A computer system for determining derivational relatedness among words, comprising:
a computer having storage;
a dictionary stored in the storage and containing semantic information for words for a derived word and a putative base word; and
a derivational computer program for determining whether the derived word derives from the putative base word by comparing the semantic information contained in the dictionary to semantic information derived from a definition of the derived word contained in the dictionary for the putative base word, the derivational computer program executed in the computer system.
19. A method in a computer system for determining the derivational relatedness of words, the computer system having a machine-readable dictionary and having semantic information for morphemes, the machine-readable dictionary having definitional information, the method comprising the steps of:
receiving a word;
parsing the word to generate a putative base word and a morpheme for the word;
identifying semantic information for the word based on the definitional information in the machine-readable dictionary;
augmenting the semantic information for the morpheme with the putative base word;
comparing the semantic information for the word and the augmented semantic information; and
indicating the derivational relatedness of the word to the putative base word based on the comparison.
20. The method of claim 19 wherein the step of comparing includes the step of comparing isomorphically related semantic information for the derived word and the morpheme augmented with the putative base word.
21. A method in a computer system for determining the derivational relatedness of words, the computer system having a machine-readable dictionary and having semantic information for morphemes, the machine-readable dictionary having definitional information, the method comprising the steps of:
receiving a word;
parsing the word to generate a putative base word and a morpheme for the word;
identifying semantic information for the word based on the definitional information in the machine-readable dictionary; and
comparing the semantic information for the word and the semantic information for the morpheme augmented with the putative base word, wherein the comparison indicates the derivational relatedness of the word to the putative base word by comparing isomorphically related semantic information for the derived word and the morpheme augmented with the putative base word and by indicating an increased likelihood of derivational relatedness when the isomorphically related semantic information has related lemma words.
22. A method in a computer system of determining whether a word is derived from a certain sense of a putative base word that is combined with a morpheme, the word and the morpheme having semantic information specifying an attribute and lemma word, the method comprising the steps of:
for each sense of the putative base word, combining the semantic information of the morpheme with the sense of the putative base word;
for each sense of the putative base word, comparing the semantic information for the word with the semantic information for the morpheme combined with that sense such that when an attribute of the semantic information for the word and an attribute of the semantic information for the morpheme are equivalent and the lemma word of that attribute for the word is related to the lemma word of that attribute for the morpheme, then a likelihood that the word derives from the putative base word is increased; and
selecting the sense with the greatest likelihood as the sense from which the word is derived.
23. A method in computer system for determining that a word is not derived from a putative base word combined with a morpheme, the method comprising the steps of:
selecting hypernym semantic information for the morpheme and the word, the hypernym semantic information having a value;
determining whether the selected hypernym semantic information is related to the value of the selected hypernym information for the word; and
when the value of the selected hypernym semantic information for the morpheme is not related to the value of the selected hypernym semantic information for the word, indicating that the word is not derived from the putative base word.
24. A computer-readable memory device for controlling a computer system to determine whether a derived word is morphologically related to a putative base word combined with a morpheme, by:
combining semantic information derived from a definition of the morpheme with the putative base word;
comparing semantic information derived from a definition of the derived word to the combined semantic information; and
determining whether the comparison indicates derivational relatedness of the derived word to the putative base word, the derivational relatedness indicating whether the derived word is morphologically related to the putative base word.
25. A computer-readable medium containing instructions for causing a computer system to determine derivational relatedness among words, the computer system having a dictionary with an entry for each head word of the dictionary, each entry having definitional information, by:
receiving a derived word, a putative base word, and a morpheme;
comparing semantic information derived from the definitional information of the derived word with semantic information derived from the definitional information of the morpheme combined with the putative base word; and
generating, based on the comparison, an indicator of the derivational relatedness of the derived word to the putative base word.
26. A computer-readable medium containing instructions for causing a computer system to determine derivational relatedness among words, the computer system having a dictionary with an entry for each head word of the dictionary, each entry having definitional information, by:
receiving a derived word, a putative base word, and a morpheme;
comparing semantic information derived from the definitional information of the derived word with semantic information derived from the definitional information of the morpheme combined with the putative base word; and
generating, based on the comparison, an indicator of the derivational relatedness of the derived word to the putative base word, wherein the generated indicator is stored in the dictionary to indicate that the derived word derives from the putative base word with an indicated degree of confidence.
27. A computer-readable medium containing instructions for causing a computer system to determine whether a word is derived from a certain sense of a putative base word that is combined with a morpheme, the word and the morpheme having semantic information specifying an attribute and lemma word, by:
for each sense of the putative base word, combining the semantic information of the morpheme with the sense of the putative base word;
for each sense of the putative base word, comparing the semantic information for the word with the semantic information for the morpheme combined with that sense such that when an attribute of the semantic information for the word and an attribute of the semantic information for the morpheme are equivalent and the lemma word of that attribute for the word is related to the lemma word of that attribute for the morpheme, then a likelihood that the word derives from the putative base word is increased; and
selecting the sense with the greatest likelihood as the sense from which the word is derived.
28. A computer-readable medium containing instructions for causing a computer system to determine that a word is not derived from a putative base word combined with a morpheme, by:
selecting hypernym semantic information for the morpheme and the word, the hypernym semantic information having a value; and
when the value of the hypernym semantic information for the morpheme is not related to the value of the hypernym semantic information for the word, indicating that the word is not derived from the putative base word.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US08/195,346 US5724594A (en) | 1994-02-10 | 1994-02-10 | Method and system for automatically identifying morphological information from a machine-readable dictionary |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US08/195,346 US5724594A (en) | 1994-02-10 | 1994-02-10 | Method and system for automatically identifying morphological information from a machine-readable dictionary |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| US5724594A true US5724594A (en) | 1998-03-03 |
Family
ID=22721073
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| US08/195,346 Expired - Lifetime US5724594A (en) | 1994-02-10 | 1994-02-10 | Method and system for automatically identifying morphological information from a machine-readable dictionary |
Country Status (1)
| Country | Link |
|---|---|
| US (1) | US5724594A (en) |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5878386A (en) * | 1996-06-28 | 1999-03-02 | Microsoft Corporation | Natural language parser with dictionary-based part-of-speech probabilities |
| US6098033A (en) * | 1997-07-31 | 2000-08-01 | Microsoft Corporation | Determining similarity between words |
| US6138085A (en) * | 1997-07-31 | 2000-10-24 | Microsoft Corporation | Inferring semantic relations |
| US6192333B1 (en) * | 1998-05-12 | 2001-02-20 | Microsoft Corporation | System for creating a dictionary |
| WO2001080077A1 (en) * | 2000-04-18 | 2001-10-25 | Korea Telecom | Method and system for retrieving information based on meaningful core word |
| US6405199B1 (en) * | 1998-10-30 | 2002-06-11 | Novell, Inc. | Method and apparatus for semantic token generation based on marked phrases in a content stream |
| US6408266B1 (en) * | 1997-04-01 | 2002-06-18 | Yeong Kaung Oon | Didactic and content oriented word processing method with incrementally changed belief system |
| US6487545B1 (en) * | 1995-05-31 | 2002-11-26 | Oracle Corporation | Methods and apparatus for classifying terminology utilizing a knowledge catalog |
| US20020194223A1 (en) * | 2000-10-16 | 2002-12-19 | Text Analysis International, Inc. | Computer programming language, system and method for building text analyzers |
| US20030023422A1 (en) * | 2001-06-01 | 2003-01-30 | Menezes Arul A. | Scaleable machine translation system |
| US20030126106A1 (en) * | 2001-12-27 | 2003-07-03 | Coen Gary A. | Database analysis tool |
| US20030191625A1 (en) * | 1999-11-05 | 2003-10-09 | Gorin Allen Louis | Method and system for creating a named entity language model |
| US6871174B1 (en) | 1997-03-07 | 2005-03-22 | Microsoft Corporation | System and method for matching a textual input to a lexical knowledge base and for utilizing results of that match |
| US20050141391A1 (en) * | 2000-07-13 | 2005-06-30 | Tetsuo Ueyama | Optical pickup |
| USRE39090E1 (en) * | 1997-07-03 | 2006-05-02 | Activeword Systems, Inc. | Semantic user interface |
| US20070011160A1 (en) * | 2005-07-07 | 2007-01-11 | Denis Ferland | Literacy automation software |
| US20080154848A1 (en) * | 2006-12-20 | 2008-06-26 | Microsoft Corporation | Search, Analysis and Comparison of Content |
| US20080177544A1 (en) * | 1999-11-05 | 2008-07-24 | At&T Corp. | Method and system for automatic detecting morphemes in a task classification system using lattices |
| US20080208566A1 (en) * | 2007-02-23 | 2008-08-28 | Microsoft Corporation | Automated word-form transformation and part of speech tag assignment |
| US20090006078A1 (en) * | 2007-06-27 | 2009-01-01 | Vladimir Selegey | Method and system for natural language dictionary generation |
| US20100010806A1 (en) * | 2008-06-27 | 2010-01-14 | Li Sha Bi | Storage system for symptom information of Traditional Chinese Medicine (TCM) and method for storing TCM symptom information |
| US7734459B2 (en) | 2001-06-01 | 2010-06-08 | Microsoft Corporation | Automatic extraction of transfer mappings from bilingual corpora |
| US8392188B1 (en) | 1999-11-05 | 2013-03-05 | At&T Intellectual Property Ii, L.P. | Method and system for building a phonotactic model for domain independent speech recognition |
| US9852125B2 (en) | 2015-09-02 | 2017-12-26 | International Business Machines Corporation | Dynamic portmanteau word semantic identification |
| US10878174B1 (en) | 2020-06-24 | 2020-12-29 | Starmind Ag | Advanced text tagging using key phrase extraction and key phrase generation |
| US11379763B1 (en) | 2021-08-10 | 2022-07-05 | Starmind Ag | Ontology-based technology platform for mapping and filtering skills, job titles, and expertise topics |
| US11487817B2 (en) * | 2017-03-28 | 2022-11-01 | Fujitsu Limited | Index generation method, data retrieval method, apparatus of index generation |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4775956A (en) * | 1984-01-30 | 1988-10-04 | Hitachi, Ltd. | Method and system for information storing and retrieval using word stems and derivative pattern codes representing familes of affixes |
| US4796185A (en) * | 1983-07-08 | 1989-01-03 | Motokazu Yoshimura | Typewriter having electronic dictionary with efficient storage, for printing any stored information associated with a entered word or entered word |
| US4959785A (en) * | 1987-05-26 | 1990-09-25 | Sharp Kabushiki Kaisha | Character processing system with spelling check function that utilizes condensed word storage and indexed retrieval |
| US5099426A (en) * | 1989-01-19 | 1992-03-24 | International Business Machines Corporation | Method for use of morphological information to cross reference keywords used for information retrieval |
| US5099425A (en) * | 1988-12-13 | 1992-03-24 | Matsushita Electric Industrial Co., Ltd. | Method and apparatus for analyzing the semantics and syntax of a sentence or a phrase |
| US5225981A (en) * | 1986-10-03 | 1993-07-06 | Ricoh Company, Ltd. | Language analyzer for morphemically and syntactically analyzing natural languages by using block analysis and composite morphemes |
| US5227971A (en) * | 1988-06-15 | 1993-07-13 | Hitachi, Ltd. | Apparatus for and method of selecting a target language equivalent of a predicate word in a source language word string in a machine translation system |
| US5243520A (en) * | 1990-08-21 | 1993-09-07 | General Electric Company | Sense discrimination system and method |
| US5251129A (en) * | 1990-08-21 | 1993-10-05 | General Electric Company | Method for automated morphological analysis of word structure |
| US5276616A (en) * | 1989-10-16 | 1994-01-04 | Sharp Kabushiki Kaisha | Apparatus for automatically generating index |
| US5311429A (en) * | 1989-05-17 | 1994-05-10 | Hitachi, Ltd. | Maintenance support method and apparatus for natural language processing system |
| US5321608A (en) * | 1990-11-30 | 1994-06-14 | Hitachi, Ltd. | Method and system for processing natural language |
| US5357430A (en) * | 1991-10-25 | 1994-10-18 | International Business Machines Corporation | System for text translation |
| US5406480A (en) * | 1992-01-17 | 1995-04-11 | Matsushita Electric Industrial Co., Ltd. | Building and updating of co-occurrence dictionary and analyzing of co-occurrence and meaning |
| US5418717A (en) * | 1990-08-27 | 1995-05-23 | Su; Keh-Yih | Multiple score language processing system |
-
1994
- 1994-02-10 US US08/195,346 patent/US5724594A/en not_active Expired - Lifetime
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4796185A (en) * | 1983-07-08 | 1989-01-03 | Motokazu Yoshimura | Typewriter having electronic dictionary with efficient storage, for printing any stored information associated with a entered word or entered word |
| US4775956A (en) * | 1984-01-30 | 1988-10-04 | Hitachi, Ltd. | Method and system for information storing and retrieval using word stems and derivative pattern codes representing familes of affixes |
| US5225981A (en) * | 1986-10-03 | 1993-07-06 | Ricoh Company, Ltd. | Language analyzer for morphemically and syntactically analyzing natural languages by using block analysis and composite morphemes |
| US4959785A (en) * | 1987-05-26 | 1990-09-25 | Sharp Kabushiki Kaisha | Character processing system with spelling check function that utilizes condensed word storage and indexed retrieval |
| US5227971A (en) * | 1988-06-15 | 1993-07-13 | Hitachi, Ltd. | Apparatus for and method of selecting a target language equivalent of a predicate word in a source language word string in a machine translation system |
| US5099425A (en) * | 1988-12-13 | 1992-03-24 | Matsushita Electric Industrial Co., Ltd. | Method and apparatus for analyzing the semantics and syntax of a sentence or a phrase |
| US5099426A (en) * | 1989-01-19 | 1992-03-24 | International Business Machines Corporation | Method for use of morphological information to cross reference keywords used for information retrieval |
| US5311429A (en) * | 1989-05-17 | 1994-05-10 | Hitachi, Ltd. | Maintenance support method and apparatus for natural language processing system |
| US5276616A (en) * | 1989-10-16 | 1994-01-04 | Sharp Kabushiki Kaisha | Apparatus for automatically generating index |
| US5243520A (en) * | 1990-08-21 | 1993-09-07 | General Electric Company | Sense discrimination system and method |
| US5251129A (en) * | 1990-08-21 | 1993-10-05 | General Electric Company | Method for automated morphological analysis of word structure |
| US5418717A (en) * | 1990-08-27 | 1995-05-23 | Su; Keh-Yih | Multiple score language processing system |
| US5321608A (en) * | 1990-11-30 | 1994-06-14 | Hitachi, Ltd. | Method and system for processing natural language |
| US5357430A (en) * | 1991-10-25 | 1994-10-18 | International Business Machines Corporation | System for text translation |
| US5406480A (en) * | 1992-01-17 | 1995-04-11 | Matsushita Electric Industrial Co., Ltd. | Building and updating of co-occurrence dictionary and analyzing of co-occurrence and meaning |
Non-Patent Citations (42)
| Title |
|---|
| "Quirk, Randolph et al., A Comprehensive Grammar of the English Language,", Longman, New York, pp. 1548-1549. |
| "Word Formation," Collins Cobuild English Guides 2, Harper Collins Publishers, The University of Birmingham, pp. 48-49. |
| Adriaens, Geert and Steven L. Small, "Word Expert Parsing Revisited in a Cognitive Science Perspective," Lexical Ambiguity Resolution: Perspectives from Psycholinguistics, Neuropsychology, and Artificial Intelligence, Morgan Kaufmann Publishers, Inc., San Mateo, California, pp. 13-43. |
| Adriaens, Geert and Steven L. Small, Word Expert Parsing Revisited in a Cognitive Science Perspective, Lexical Ambiguity Resolution: Perspectives from Psycholinguistics, Neuropsychology, and Artificial Intelligence, Morgan Kaufmann Publishers, Inc., San Mateo, California, pp. 13 43. * |
| Ahlswede, Thomas and Martha Evens, "Parsing vs. Text Processing in the Analysis of Dictionary Definitions," in Proceedings of the Conference for the 26th Annual Meeting of the Association for Computational Linguistics, State University of New York at Buffalo, New York, Jun. 7-10, 1988, pp. 217-224. |
| Ahlswede, Thomas and Martha Evens, Parsing vs. Text Processing in the Analysis of Dictionary Definitions, in Proceedings of the Conference for the 26th Annual Meeting of the Association for Computational Linguistics, State University of New York at Buffalo, New York, Jun. 7 10, 1988, pp. 217 224. * |
| Amsler, Robert A., "The Structure of the Merriam-Webster Pocket Dictionary," dissertation, University of Texas at Austin, 1980, pp. iv-v, 1-5, 130-131. |
| Amsler, Robert A., The Structure of the Merriam Webster Pocket Dictionary, dissertation, University of Texas at Austin, 1980, pp. iv v, 1 5, 130 131. * |
| Binot, Jean Louis and Karen Jensen, A Semantic Expert Using an Online Standard Dictionary, in Jensen et al. (eds.), Natural Language Processing: The PLNLP Approach, Kluwer Academic Publishers, Boston, 1987, pp. 135 147. * |
| Binot, Jean-Louis and Karen Jensen, "A Semantic Expert Using an Online Standard Dictionary," in Jensen et al. (eds.), Natural Language Processing: The PLNLP Approach, Kluwer Academic Publishers, Boston, 1987, pp. 135-147. |
| Calzolari, Nicoletta, "The dictionary and the thesaurus can be combined," in Evens (ed.) Relational models of the lexicon: Representing knowledge in semantic networks; Chapter 3, Cambridge University Press, pp. 75-96. |
| Calzolari, Nicoletta, The dictionary and the thesaurus can be combined, in Evens (ed.) Relational models of the lexicon: Representing knowledge in semantic networks; Chapter 3, Cambridge University Press, pp. 75 96. * |
| Carroll, John and Claire Grover, "The derivation of a large computational lexicon for English from LDOCE, in Boguraev and Briscoe (eds.), " Computational Lexicography for Natural Language Processing, Longman, New York, pp. 117-183. |
| Carroll, John and Claire Grover, The derivation of a large computational lexicon for English from LDOCE, in Boguraev and Briscoe (eds.), Computational Lexicography for Natural Language Processing, Longman, New York, pp. 117 183. * |
| Chodorow, Martin S.et al., "Extracting Semantic Hierarchies from a Large On-Line Dictionary," in Proceedings of the Conference for the 23rd Annual Meeting of the Association for Computational, University of Chicago, Illinois, Jul. 8-12, 1985, pp. 299-304. |
| Chodorow, Martin S.et al., Extracting Semantic Hierarchies from a Large On Line Dictionary, in Proceedings of the Conference for the 23rd Annual Meeting of the Association for Computational, University of Chicago, Illinois, Jul. 8 12, 1985, pp. 299 304. * |
| Dolan, William et al., "Automatically Deriving Structured Knowledge Bases from On-Line Dictionaries," Proceedings of the Conference for The First Conference of the Pacific Association for Computational Linguistics, The Harbour Centre Campus of Simon Fraser University, Vancouver, B.C., Apr. 21-24, 1993, Columbia, pp. 5-14. |
| Dolan, William et al., Automatically Deriving Structured Knowledge Bases from On Line Dictionaries, Proceedings of the Conference for The First Conference of the Pacific Association for Computational Linguistics, The Harbour Centre Campus of Simon Fraser University, Vancouver, B.C., Apr. 21 24, 1993, Columbia, pp. 5 14. * |
| Dumitrescu, Cristian, "Lexicon Design Using a Paradigmatic Approach," in Proceedings of the Conference for the Third Conference on Applied Natural Language Processing, Trento, Italy, Association for Computational Linguistics, Mar. 31-Apr. 3, 1993, pp. 247-248. |
| Dumitrescu, Cristian, Lexicon Design Using a Paradigmatic Approach, in Proceedings of the Conference for the Third Conference on Applied Natural Language Processing, Trento, Italy, Association for Computational Linguistics, Mar. 31 Apr. 3, 1993, pp. 247 248. * |
| Krovetz, Robert, "Viewing Morphology as an Inference Process," in Proceedings of the ACM-SIGIR Conference on Research and Development in Information Retrieval, 1993, pp. 191-202. |
| Krovetz, Robert, Viewing Morphology as an Inference Process, in Proceedings of the ACM SIGIR Conference on Research and Development in Information Retrieval, 1993, pp. 191 202. * |
| Marchand, Hans, The Categories and Types of Present Day English Word Formation: A Synchronic Diachronic Approach, C.H. Beck sche Verlagsbuchhandlung, Munich, Germany, pp. 273 281. * |
| Marchand, Hans, The Categories and Types of Present-Day English Word-Formation: A Synchronic-Diachronic Approach, C.H. Beck'sche Verlagsbuchhandlung, Munich, Germany, pp. 273-281. |
| Markowitz, Judith et al., "Semantically Significant Patterns in Dictionary Definitions,"in Proceedings Of the Conference for the 24th Annual Meeting of the Association for Computational Linguistics, Columbia University, New York, Jun. 10-13, 1986, pp. 112-119. |
| Markowitz, Judith et al., Semantically Significant Patterns in Dictionary Definitions, in Proceedings Of the Conference for the 24th Annual Meeting of the Association for Computational Linguistics, Columbia University, New York, Jun. 10 13, 1986, pp. 112 119. * |
| Montemagni, Simonetta and Lucy Vanderwende, "Structural Patterns vs. String Patterns for Extracting Semantic Information from Dictionaries," in Proceedings of the Fifteenth International Conference on Computational Linguistics, Nantes, Aug. 23-28, 1992, pp. 546-553. |
| Montemagni, Simonetta and Lucy Vanderwende, Structural Patterns vs. String Patterns for Extracting Semantic Information from Dictionaries, in Proceedings of the Fifteenth International Conference on Computational Linguistics, Nantes, Aug. 23 28, 1992, pp. 546 553. * |
| Neff, Mary S.and Bravamir K. Bogaraev, "Dictionaries, Dictionary Grammars and Dictionary Entry Parsing," in Proceedings of the Conference for the 27th Annual Meeting of the Association for Linguistics, 1989, University of Bristish Columbia, Vancouver, B.C., Jun. 26-19, 1989, pp. 91-101. |
| Neff, Mary S.and Bravamir K. Bogaraev, Dictionaries, Dictionary Grammars and Dictionary Entry Parsing, in Proceedings of the Conference for the 27th Annual Meeting of the Association for Linguistics, 1989, University of Bristish Columbia, Vancouver, B.C., Jun. 26 19, 1989, pp. 91 101. * |
| Pentheroudakis, Joseph E. and Dan W. Higinbotham, "MORFOGEN--A Morphology Grammar Builder and Dictionary Interface Tool," presented at the 1991 meeting of the Deseret Language and Linguistics Society, Brigham Young University, Utah, pp. 1-19. |
| Pentheroudakis, Joseph E. and Dan W. Higinbotham, MORFOGEN A Morphology Grammar Builder and Dictionary Interface Tool, presented at the 1991 meeting of the Deseret Language and Linguistics Society, Brigham Young University, Utah, pp. 1 19. * |
| Proctor, Paul (ed.), Longman Dictionary of Contemporary English, Longman Group Limited, United Kingdom, 1978, pp. xxvi xxvii. * |
| Proctor, Paul (ed.), Longman Dictionary of Contemporary English, Longman Group Limited, United Kingdom, 1978, pp. xxvi-xxvii. |
| Quirk, Randolph et al., A Comprehensive Grammar of the English Language, , Longman, New York, pp. 1548 1549. * |
| Russell, G.J. et al., "A Dictionary and Morphological Analyser for English," in Proceedings for the 11th International Conference on Computational Linguistics, Aug. 25-29, 1986, pp. 277-279. |
| Russell, G.J. et al., A Dictionary and Morphological Analyser for English, in Proceedings for the 11th International Conference on Computational Linguistics, Aug. 25 29, 1986, pp. 277 279. * |
| Sproat, Richard, "Computational Morphology," Morphology and Computation, The MIT Press, Cambridge, Mass., pp. 124-131. |
| Sproat, Richard, Computational Morphology, Morphology and Computation, The MIT Press, Cambridge, Mass., pp. 124 131. * |
| Strzalkowski, Tomek and Barbara Vauthey, "Information Retrieval Using Robust Natural Language Processing," in Proceedings of the Conference for the 30th Annual Meeting of the Association for Computational Linguistics, Jun. 28-Jul. 2, 1992, University of Delaware, Delaware, pp. 104-110. |
| Strzalkowski, Tomek and Barbara Vauthey, Information Retrieval Using Robust Natural Language Processing, in Proceedings of the Conference for the 30th Annual Meeting of the Association for Computational Linguistics, Jun. 28 Jul. 2, 1992, University of Delaware, Delaware, pp. 104 110. * |
| Word Formation, Collins Cobuild English Guides 2, Harper Collins Publishers, The University of Birmingham, pp. 48 49. * |
Cited By (48)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6487545B1 (en) * | 1995-05-31 | 2002-11-26 | Oracle Corporation | Methods and apparatus for classifying terminology utilizing a knowledge catalog |
| US5878386A (en) * | 1996-06-28 | 1999-03-02 | Microsoft Corporation | Natural language parser with dictionary-based part-of-speech probabilities |
| US6871174B1 (en) | 1997-03-07 | 2005-03-22 | Microsoft Corporation | System and method for matching a textual input to a lexical knowledge base and for utilizing results of that match |
| US7013264B2 (en) | 1997-03-07 | 2006-03-14 | Microsoft Corporation | System and method for matching a textual input to a lexical knowledge based and for utilizing results of that match |
| US20050065777A1 (en) * | 1997-03-07 | 2005-03-24 | Microsoft Corporation | System and method for matching a textual input to a lexical knowledge based and for utilizing results of that match |
| US6408266B1 (en) * | 1997-04-01 | 2002-06-18 | Yeong Kaung Oon | Didactic and content oriented word processing method with incrementally changed belief system |
| USRE39090E1 (en) * | 1997-07-03 | 2006-05-02 | Activeword Systems, Inc. | Semantic user interface |
| US6098033A (en) * | 1997-07-31 | 2000-08-01 | Microsoft Corporation | Determining similarity between words |
| US6138085A (en) * | 1997-07-31 | 2000-10-24 | Microsoft Corporation | Inferring semantic relations |
| US6192333B1 (en) * | 1998-05-12 | 2001-02-20 | Microsoft Corporation | System for creating a dictionary |
| US6405199B1 (en) * | 1998-10-30 | 2002-06-11 | Novell, Inc. | Method and apparatus for semantic token generation based on marked phrases in a content stream |
| US8392188B1 (en) | 1999-11-05 | 2013-03-05 | At&T Intellectual Property Ii, L.P. | Method and system for building a phonotactic model for domain independent speech recognition |
| US20030191625A1 (en) * | 1999-11-05 | 2003-10-09 | Gorin Allen Louis | Method and system for creating a named entity language model |
| US9514126B2 (en) | 1999-11-05 | 2016-12-06 | At&T Intellectual Property Ii, L.P. | Method and system for automatically detecting morphemes in a task classification system using lattices |
| US20080288244A1 (en) * | 1999-11-05 | 2008-11-20 | At&T Corp. | Method and system for automatically detecting morphemes in a task classification system using lattices |
| US8909529B2 (en) | 1999-11-05 | 2014-12-09 | At&T Intellectual Property Ii, L.P. | Method and system for automatically detecting morphemes in a task classification system using lattices |
| US8612212B2 (en) | 1999-11-05 | 2013-12-17 | At&T Intellectual Property Ii, L.P. | Method and system for automatically detecting morphemes in a task classification system using lattices |
| US20080177544A1 (en) * | 1999-11-05 | 2008-07-24 | At&T Corp. | Method and system for automatic detecting morphemes in a task classification system using lattices |
| US20080215328A1 (en) * | 1999-11-05 | 2008-09-04 | At&T Corp. | Method and system for automatically detecting morphemes in a task classification system using lattices |
| US8200491B2 (en) | 1999-11-05 | 2012-06-12 | At&T Intellectual Property Ii, L.P. | Method and system for automatically detecting morphemes in a task classification system using lattices |
| US8010361B2 (en) | 1999-11-05 | 2011-08-30 | At&T Intellectual Property Ii, L.P. | Method and system for automatically detecting morphemes in a task classification system using lattices |
| WO2001080077A1 (en) * | 2000-04-18 | 2001-10-25 | Korea Telecom | Method and system for retrieving information based on meaningful core word |
| US20090144249A1 (en) * | 2000-04-18 | 2009-06-04 | Kt Corporation | Method and system for retrieving information based on meaningful core word |
| US20050141391A1 (en) * | 2000-07-13 | 2005-06-30 | Tetsuo Ueyama | Optical pickup |
| US20020194223A1 (en) * | 2000-10-16 | 2002-12-19 | Text Analysis International, Inc. | Computer programming language, system and method for building text analyzers |
| US7734459B2 (en) | 2001-06-01 | 2010-06-08 | Microsoft Corporation | Automatic extraction of transfer mappings from bilingual corpora |
| US7050964B2 (en) | 2001-06-01 | 2006-05-23 | Microsoft Corporation | Scaleable machine translation system |
| US20030023422A1 (en) * | 2001-06-01 | 2003-01-30 | Menezes Arul A. | Scaleable machine translation system |
| US20060085180A1 (en) * | 2001-06-01 | 2006-04-20 | Microsoft Corporation | Scaleable machine translation |
| US8275605B2 (en) | 2001-06-01 | 2012-09-25 | Microsoft Corporation | Machine language translation with transfer mappings having varying context |
| US7206735B2 (en) | 2001-06-01 | 2007-04-17 | Microsoft Corporation | Scaleable machine translation |
| US20100223049A1 (en) * | 2001-06-01 | 2010-09-02 | Microsoft Corporation | Machine language translation with transfer mappings having varying context |
| US7165058B2 (en) * | 2001-12-27 | 2007-01-16 | The Boeing Company | Database analysis tool |
| US20030126106A1 (en) * | 2001-12-27 | 2003-07-03 | Coen Gary A. | Database analysis tool |
| US20070011160A1 (en) * | 2005-07-07 | 2007-01-11 | Denis Ferland | Literacy automation software |
| US8065307B2 (en) | 2006-12-20 | 2011-11-22 | Microsoft Corporation | Parsing, analysis and scoring of document content |
| US20080154848A1 (en) * | 2006-12-20 | 2008-06-26 | Microsoft Corporation | Search, Analysis and Comparison of Content |
| US20080208566A1 (en) * | 2007-02-23 | 2008-08-28 | Microsoft Corporation | Automated word-form transformation and part of speech tag assignment |
| US20090006078A1 (en) * | 2007-06-27 | 2009-01-01 | Vladimir Selegey | Method and system for natural language dictionary generation |
| US8849651B2 (en) | 2007-06-27 | 2014-09-30 | Abbyy Infopoisk Llc | Method and system for natural language dictionary generation |
| US8812296B2 (en) | 2007-06-27 | 2014-08-19 | Abbyy Infopoisk Llc | Method and system for natural language dictionary generation |
| US20100010806A1 (en) * | 2008-06-27 | 2010-01-14 | Li Sha Bi | Storage system for symptom information of Traditional Chinese Medicine (TCM) and method for storing TCM symptom information |
| US9852125B2 (en) | 2015-09-02 | 2017-12-26 | International Business Machines Corporation | Dynamic portmanteau word semantic identification |
| US9852124B2 (en) | 2015-09-02 | 2017-12-26 | International Business Machines Corporation | Dynamic portmanteau word semantic identification |
| US10108602B2 (en) | 2015-09-02 | 2018-10-23 | International Business Machines Corporation | Dynamic portmanteau word semantic identification |
| US11487817B2 (en) * | 2017-03-28 | 2022-11-01 | Fujitsu Limited | Index generation method, data retrieval method, apparatus of index generation |
| US10878174B1 (en) | 2020-06-24 | 2020-12-29 | Starmind Ag | Advanced text tagging using key phrase extraction and key phrase generation |
| US11379763B1 (en) | 2021-08-10 | 2022-07-05 | Starmind Ag | Ontology-based technology platform for mapping and filtering skills, job titles, and expertise topics |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US5724594A (en) | Method and system for automatically identifying morphological information from a machine-readable dictionary | |
| Hu et al. | Ella: Equip diffusion models with llm for enhanced semantic alignment | |
| Rohde et al. | An improved model of semantic similarity based on lexical co-occurrence | |
| US5995922A (en) | Identifying information related to an input word in an electronic dictionary | |
| US7016828B1 (en) | Text-to-scene conversion | |
| Harabagiu et al. | Wordnet 2-a morphologically and semantically enhanced resource | |
| US6487545B1 (en) | Methods and apparatus for classifying terminology utilizing a knowledge catalog | |
| US8086028B2 (en) | Text-to-scene conversion | |
| US4674066A (en) | Textual database system using skeletonization and phonetic replacement to retrieve words matching or similar to query words | |
| US6115683A (en) | Automatic essay scoring system using content-based techniques | |
| US8478581B2 (en) | Interlingua, interlingua engine, and interlingua machine translation system | |
| US20230138014A1 (en) | System and method for performing a search in a vector space based search engine | |
| Rohde et al. | An improved method for deriving word meaning from lexical co-occurrence | |
| US20040093567A1 (en) | Spelling and grammar checking system | |
| US20080275694A1 (en) | Method and system for automatically extracting relations between concepts included in text | |
| WO1998039711A1 (en) | Language analysis system and method | |
| Meyers | Annotation Guidelines for NomBank–Noun Argument Structure for PropBank 2007 | |
| Evens et al. | A lexicon for a computer question-answering system | |
| JP2007047974A (en) | Information extraction apparatus and information extraction method | |
| JP4572321B2 (en) | Document output apparatus and control method thereof | |
| Freihat | An organizational approach to the polysemy problem in wordnet | |
| KR102844089B1 (en) | Device and Method for Converting Prompt for Generating Image | |
| JP4336404B2 (en) | Information retrieval apparatus and method, and computer-readable memory | |
| Way | Conceptual graph overview | |
| JP4336403B2 (en) | Information retrieval apparatus and method, and computer-readable memory |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| AS | Assignment |
Owner name: MICROSOFT CORPORATION, WASHINGTON Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:PENTHEROUDAKIS, JOSEPH E.;REEL/FRAME:006881/0352 Effective date: 19940209 |
|
| STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
| FPAY | Fee payment |
Year of fee payment: 4 |
|
| FPAY | Fee payment |
Year of fee payment: 8 |
|
| FPAY | Fee payment |
Year of fee payment: 12 |
|
| AS | Assignment |
Owner name: MICROSOFT TECHNOLOGY LICENSING, LLC, WASHINGTON Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MICROSOFT CORPORATION;REEL/FRAME:034541/0001 Effective date: 20141014 |